arXiv subjects

Richard Peng

Publications and source records attributed to Richard Peng.

90 records · Page 5Linked to original sources

arXiv2014

A Generalized Cheeger Inequality

The generalized conductance $ϕ(G,H)$ between two graphs $G$ and $H$ on the same vertex set $V$ is defined as the ratio $$ ϕ(G,H) = \min_{S\subseteq V} \frac{cap_G(S,\bar{S})}{ cap_H(S,\bar{S})}, $$ where $cap_G(S,\bar{S})$ is the total weight of the edges crossing from $S$ to $\bar{S}=V-S$. We show that the minimum generalized eigenvalue $λ(L_G,L_H)$ of the pair of Laplacians $L_G$ and $L_H$ satisfies $$ λ(L_G,L_H) \geq ϕ(G,H) ϕ(G)/8, $$ where $ϕ(G)$ is the usual conductance of $G$. A generalized cut that meets this bound can be obtained from the generalized eigenvector corresponding to $λ(L_G,L_H)$. The inequality complements a recent proof that $ϕ(G)$ cannot be replaced by $Θ(ϕ(G,H))$ in the above inequality, unless the Unique Games Conjecture is false.

cs.DM↗

arXiv2014

Scalable Parallel Factorizations of SDD Matrices and Efficient Sampling for Gaussian Graphical Models

Motivated by a sampling problem basic to computational statistical inference, we develop a nearly optimal algorithm for a fundamental problem in spectral graph theory and numerical analysis. Given an $n\times n$ SDDM matrix ${\bf \mathbf{M}}$, and a constant $-1 \leq p \leq 1$, our algorithm gives efficient access to a sparse $n\times n$ linear operator $\tilde{\mathbf{C}}$ such that $${\mathbf{M}}^{p} \approx \tilde{\mathbf{C}} \tilde{\mathbf{C}}^\top.$$ The solution is based on factoring ${\bf \mathbf{M}}$ into a product of simple and sparse matrices using squaring and spectral sparsification. For ${\mathbf{M}}$ with $m$ non-zero entries, our algorithm takes work nearly-linear in $m$, and polylogarithmic depth on a parallel machine with $m$ processors. This gives the first sampling algorithm that only requires nearly linear work and $n$ i.i.d. random univariate Gaussian samples to generate i.i.d. random samples for $n$-dimensional Gaussian random fields with SDDM precision matrices. For sampling this natural subclass of Gaussian random fields, it is optimal in the randomness and nearly optimal in the work and parallel complexity. In addition, our sampling algorithm can be directly extended to Gaussian random fields with SDD precision matrices.

cs.DS↗

arXiv2014

Uniform Sampling for Matrix Approximation

Random sampling has become a critical tool in solving massive matrix problems. For linear regression, a small, manageable set of data rows can be randomly selected to approximate a tall, skinny data matrix, improving processing time significantly. For theoretical performance guarantees, each row must be sampled with probability proportional to its statistical leverage score. Unfortunately, leverage scores are difficult to compute. A simple alternative is to sample rows uniformly at random. While this often works, uniform sampling will eliminate critical row information for many natural instances. We take a fresh look at uniform sampling by examining what information it does preserve. Specifically, we show that uniform sampling yields a matrix that, in some sense, well approximates a large fraction of the original. While this weak form of approximation is not enough for solving linear regression directly, it is enough to compute a better approximation. This observation leads to simple iterative row sampling algorithms for matrix approximation that run in input-sparsity time and preserve row structure and sparsity at all intermediate steps. In addition to an improved understanding of uniform sampling, our main proof introduces a structural result of independent interest: we show that every matrix can be made to have low coherence by reweighting a small subset of its rows.

cs.DS↗

arXiv2014

Stretching Stretch

We give a generalized definition of stretch that simplifies the efficient construction of low-stretch embeddings suitable for graph algorithms. The generalization, based on discounting highly stretched edges by taking their $p$-th power for some $0 < p < 1$, is directly related to performances of existing algorithms. This discounting of high-stretch edges allows us to treat many classes of edges with coarser granularity. It leads to a two-pass approach that combines bottom-up clustering and top-down decompositions to construct these embeddings in $\mathcal{O}(m\log\log{n})$ time. Our algorithm parallelizes readily and can also produce generalizations of low-stretch subgraphs.

cs.DS↗

arXiv2014

Preconditioning in Expectation

We show that preconditioners constructed by random sampling can perform well without meeting the standard requirements of iterative methods. When applied to graph Laplacians, this leads to ultra-sparsifiers that in expectation behave as the nearly-optimal ones given by [Kolla-Makarychev-Saberi-Teng STOC`10]. Combining this with the recursive preconditioning framework by [Spielman-Teng STOC`04] and improved embedding algorithms, this leads to algorithms that solve symmetric diagonally dominant linear systems and electrical flow problems in expected time close to $m\log^{1/2}n$ .

cs.DS↗

arXiv2013

Faster spectral sparsification and numerical algorithms for SDD matrices

We study algorithms for spectral graph sparsification. The input is a graph $G$ with $n$ vertices and $m$ edges, and the output is a sparse graph $\tilde{G}$ that approximates $G$ in an algebraic sense. Concretely, for all vectors $x$ and any $ε>0$, $\tilde{G}$ satisfies $$ (1-ε) x^T L_G x \leq x^T L_{\tilde{G}} x \leq (1+ε) x^T L_G x, $$ where $L_G$ and $L_{\tilde{G}}$ are the Laplacians of $G$ and $\tilde{G}$ respectively. We show that the fastest known algorithm for computing a sparsifier with $O(n\log n/ε^2)$ edges can actually run in $\tilde{O}(m\log^2 n)$ time, an $O(\log n)$ factor faster than before. We also present faster sparsification algorithms for slightly dense graphs. Specifically, we give an algorithm that runs in $\tilde{O}(m\log n)$ time and generates a sparsifier with $\tilde{O}(n\log^3{n}/ε^2)$ edges. This implies that a sparsifier with $O(n\log n/ε^2)$ edges can be computed in $\tilde{O}(m\log n)$ time for graphs with more than $O(n\log^4 n)$ edges. We also give an $\tilde{O}(m)$ time algorithm for graphs with more than $n\log^5 n (\log \log n)^3$ edges of polynomially bounded weights, and an $O(m)$ algorithm for unweighted graphs with more than $n\log^8 n (\log \log n)^3 $ edges and $n\log^{10} n (\log \log n)^5$ edges in the weighted case. The improved sparsification algorithms are employed to accelerate linear system solvers and algorithms for computing fundamental eigenvectors of slightly dense SDD matrices.

cs.DS↗

arXiv2013

An Efficient Parallel Solver for SDD Linear Systems

We present the first parallel algorithm for solving systems of linear equations in symmetric, diagonally dominant (SDD) matrices that runs in polylogarithmic time and nearly-linear work. The heart of our algorithm is a construction of a sparse approximate inverse chain for the input matrix: a sequence of sparse matrices whose product approximates its inverse. Whereas other fast algorithms for solving systems of equations in SDD matrices exploit low-stretch spanning trees, our algorithm only requires spectral graph sparsifiers.

math.NA↗

arXiv2013

Parallel Graph Decompositions Using Random Shifts

We show an improved parallel algorithm for decomposing an undirected unweighted graph into small diameter pieces with a small fraction of the edges in between. These decompositions form critical subroutines in a number of graph algorithms. Our algorithm builds upon the shifted shortest path approach introduced in [Blelloch, Gupta, Koutis, Miller, Peng, Tangwongsan, SPAA 2011]. By combining various stages of the previous algorithm, we obtain a significantly simpler algorithm with the same asymptotic guarantees as the best sequential algorithm.

cs.DS↗

arXiv2013

Fully Dynamic $(1+ε)$-Approximate Matchings

We present the first data structures that maintain near optimal maximum cardinality and maximum weighted matchings on sparse graphs in sublinear time per update. Our main result is a data structure that maintains a $(1+ε)$ approximation of maximum matching under edge insertions/deletions in worst case $O(\sqrt{m}ε^{-2})$ time per update. This improves the 3/2 approximation given in [Neiman,Solomon,STOC 2013] which runs in similar time. The result is based on two ideas. The first is to re-run a static algorithm after a chosen number of updates to ensure approximation guarantees. The second is to judiciously trim the graph to a smaller equivalent one whenever possible. We also study extensions of our approach to the weighted setting, and combine it with known frameworks to obtain arbitrary approximation ratios. For a constant $ε$ and for graphs with edge weights between 1 and N, we design an algorithm that maintains an $(1+ε)$-approximate maximum weighted matching in $O(\sqrt{m} \log N)$ time per update. The only previous result for maintaining weighted matchings on dynamic graphs has an approximation ratio of 4.9108, and was shown in [Anand,Baswana,Gupta,Sen, FSTTCS 2012, arXiv 2012].

cs.DS↗

arXiv2013

Iterative Row Sampling

There has been significant interest and progress recently in algorithms that solve regression problems involving tall and thin matrices in input sparsity time. These algorithms find shorter equivalent of a n*d matrix where n >> d, which allows one to solve a poly(d) sized problem instead. In practice, the best performances are often obtained by invoking these routines in an iterative fashion. We show these iterative methods can be adapted to give theoretical guarantees comparable and better than the current state of the art. Our approaches are based on computing the importances of the rows, known as leverage scores, in an iterative manner. We show that alternating between computing a short matrix estimate and finding more accurate approximate leverage scores leads to a series of geometrically smaller instances. This gives an algorithm that runs in $O(nnz(A) + d^{ω+ θ} ε^{-2})$ time for any $θ> 0$, where the $d^{ω+ θ}$ term is comparable to the cost of solving a regression problem on the small approximation. Our results are built upon the close connection between randomized matrix algorithms, iterative methods, and graph sparsification.

cs.DS↗

arXiv2012

Approximate Maximum Flow on Separable Undirected Graphs

We present faster algorithms for approximate maximum flow in undirected graphs with good separator structures, such as bounded genus, minor free, and geometric graphs. Given such a graph with $n$ vertices, $m$ edges along with a recursive $\sqrt{n}$-vertex separator structure, our algorithm finds an $1-ε$ approximate maximum flow in time $\tilde{O}(m^{6/5} \poly{ε^{-1}})$, ignoring poly-logarithmic terms. Similar speedups are also achieved for separable graphs with larger size separators albeit with larger run times. These bounds also apply to image problems in two and three dimensions. Key to our algorithm is an intermediate problem that we term grouped $L_2$ flow, which exists between maximum flows and electrical flows. Our algorithm also makes use of spectral vertex sparsifiers in order to remove vertices while preserving the energy dissipation of electrical flows. We also give faster spectral vertex sparsification algorithms on well separated graphs, which may be of independent interest.

cs.DS↗

arXiv2012

Runtime Guarantees for Regression Problems

We study theoretical runtime guarantees for a class of optimization problems that occur in a wide variety of inference problems. these problems are motivated by the lasso framework and have applications in machine learning and computer vision. Our work shows a close connection between these problems and core questions in algorithmic graph theory. While this connection demonstrates the difficulties of obtaining runtime guarantees, it also suggests an approach of using techniques originally developed for graph algorithms. We then show that most of these problems can be formulated as a grouped least squares problem, and give efficient algorithms for this formulation. Our algorithms rely on routines for solving quadratic minimization problems, which in turn are equivalent to solving linear systems. Finally we present some experimental results on applying our approximation algorithm to image processing problems.

cs.DS↗

arXiv2012

Faster Approximate Multicommodity Flow Using Quadratically Coupled Flows

The maximum multicommodity flow problem is a natural generalization of the maximum flow problem to route multiple distinct flows. Obtaining a $1-ε$ approximation to the multicommodity flow problem on graphs is a well-studied problem. In this paper we present an adaptation of recent advances in single-commodity flow algorithms to this problem. As the underlying linear systems in the electrical problems of multicommodity flow problems are no longer Laplacians, our approach is tailored to generate specialized systems which can be preconditioned and solved efficiently using Laplacians. Given an undirected graph with m edges and k commodities, we give algorithms that find $1-ε$ approximate solutions to the maximum concurrent flow problem and the maximum weighted multicommodity flow problem in time $\tilde{O}(m^{4/3}\poly(k,ε^{-1}))$.

cs.DS↗

arXiv2011

Near Linear-Work Parallel SDD Solvers, Low-Diameter Decomposition, and Low-Stretch Subgraphs

We present the design and analysis of a near linear-work parallel algorithm for solving symmetric diagonally dominant (SDD) linear systems. On input of a SDD $n$-by-$n$ matrix $A$ with $m$ non-zero entries and a vector $b$, our algorithm computes a vector $\tilde{x}$ such that $\norm[A]{\tilde{x} - A^+b} \leq \vareps \cdot \norm[A]{A^+b}$ in $O(m\log^{O(1)}{n}\log{\frac1ε})$ work and $O(m^{1/3+θ}\log \frac1ε)$ depth for any fixed $θ> 0$. The algorithm relies on a parallel algorithm for generating low-stretch spanning trees or spanning subgraphs. To this end, we first develop a parallel decomposition algorithm that in polylogarithmic depth and $\otilde(|E|)$ work, partitions a graph into components with polylogarithmic diameter such that only a small fraction of the original edges are between the components. This can be used to generate low-stretch spanning trees with average stretch $O(n^α)$ in $O(n^{1+α})$ work and $O(n^α)$ depth. Alternatively, it can be used to generate spanning subgraphs with polylogarithmic average stretch in $\otilde(|E|)$ work and polylogarithmic depth. We apply this subgraph construction to derive a parallel linear system solver. By using this solver in known applications, our results imply improved parallel randomized algorithms for several problems, including single-source shortest paths, maximum flow, minimum-cost flow, and approximate maximum flow.

cs.DS↗

arXiv2011

A nearly-mlogn time solver for SDD linear systems

We present an improved algorithm for solving symmetrically diagonally dominant linear systems. On input of an $n\times n$ symmetric diagonally dominant matrix $A$ with $m$ non-zero entries and a vector $b$ such that $A\bar{x} = b$ for some (unknown) vector $\bar{x}$, our algorithm computes a vector $x$ such that $||{x}-\bar{x}||_A < ε||\bar{x}||_A $ {$||\cdot||_A$ denotes the A-norm} in time $${\tilde O}(m\log n \log (1/ε)).$$ The solver utilizes in a standard way a `preconditioning' chain of progressively sparser graphs. To claim the faster running time we make a two-fold improvement in the algorithm for constructing the chain. The new chain exploits previously unknown properties of the graph sparsification algorithm given in [Koutis,Miller,Peng, FOCS 2010], allowing for stronger preconditioning properties. We also present an algorithm of independent interest that constructs nearly-tight low-stretch spanning trees in time $\tilde{O}(m\log{n})$, a factor of $O(\log{n})$ faster than the algorithm in [Abraham,Bartal,Neiman, FOCS 2008]. This speedup directly reflects on the construction time of the preconditioning chain.

cs.DS↗

arXiv2010

Efficient Triangle Counting in Large Graphs via Degree-based Vertex Partitioning

The number of triangles is a computationally expensive graph statistic which is frequently used in complex network analysis (e.g., transitivity ratio), in various random graph models (e.g., exponential random graph model) and in important real world applications such as spam detection, uncovering of the hidden thematic structure of the Web and link recommendation. Counting triangles in graphs with millions and billions of edges requires algorithms which run fast, use small amount of space, provide accurate estimates of the number of triangles and preferably are parallelizable. In this paper we present an efficient triangle counting algorithm which can be adapted to the semistreaming model. The key idea of our algorithm is to combine the sampling algorithm of Tsourakakis et al. and the partitioning of the set of vertices into a high degree and a low degree subset respectively as in the Alon, Yuster and Zwick work treating each set appropriately. We obtain a running time $O \left(m + \frac{m^{3/2} Δ\log{n}}{t ε^2} \right)$ and an $ε$ approximation (multiplicative error), where $n$ is the number of vertices, $m$ the number of edges and $Δ$ the maximum number of triangles an edge is contained. Furthermore, we show how this algorithm can be adapted to the semistreaming model with space usage $O\left(m^{1/2}\log{n} + \frac{m^{3/2} Δ\log{n}}{t ε^2} \right)$ and a constant number of passes (three) over the graph stream. We apply our methods in various networks with several millions of edges and we obtain excellent results. Finally, we propose a random projection based method for triangle counting and provide a sufficient condition to obtain an estimate with low variance.

cs.DS↗

arXiv2010

Approaching optimality for solving SDD systems

We present an algorithm that on input of an $n$-vertex $m$-edge weighted graph $G$ and a value $k$, produces an {\em incremental sparsifier} $\hat{G}$ with $n-1 + m/k$ edges, such that the condition number of $G$ with $\hat{G}$ is bounded above by $\tilde{O}(k\log^2 n)$, with probability $1-p$. The algorithm runs in time $$\tilde{O}((m \log{n} + n\log^2{n})\log(1/p)).$$ As a result, we obtain an algorithm that on input of an $n\times n$ symmetric diagonally dominant matrix $A$ with $m$ non-zero entries and a vector $b$, computes a vector ${x}$ satisfying $||{x}-A^{+}b||_A<ε||A^{+}b||_A $, in expected time $$\tilde{O}(m\log^2{n}\log(1/ε)).$$ The solver is based on repeated applications of the incremental sparsifier that produces a chain of graphs which is then used as input to a recursive preconditioned Chebyshev iteration.

cs.DS↗

arXiv2010

Approximate Dynamic Programming using Halfspace Queries and Multiscale Monge decomposition

Let $P=(P_1, P_2, \ldots, P_n)$, $P_i \in \field{R}$ for all $i$, be a signal and let $C$ be a constant. In this work our goal is to find a function $F:[n]\rightarrow \field{R}$ which optimizes the following objective function: $$ \min_{F} \sum_{i=1}^n (P_i-F_i)^2 + C\times |\{i:F_i \neq F_{i+1} \} | $$ The above optimization problem reduces to solving the following recurrence, which can be done efficiently using dynamic programming in $O(n^2)$ time: $$ OPT_i = \min_{0 \leq j \leq i-1} [ OPT_j + \sum_{k=j+1}^i (P_k - (\sum_{m=j+1}^i P_m)/(i-j) )^2 ]+ C $$ The above recurrence arises naturally in applications where we wish to approximate the original signal $P$ with another signal $F$ which consists ideally of few piecewise constant segments. Such applications include database (e.g., histogram construction), speech recognition, biology (e.g., denoising aCGH data) applications and many more. In this work we present two new techniques for optimizing dynamic programming that can handle cost functions not treated by other standard methods. The basis of our first algorithm is the definition of a constant-shifted variant of the objective function that can be efficiently approximated using state of the art methods for range searching. Our technique approximates the optimal value of our objective function within additive $ε$ error and runs in $\tilde{O}(n^{1.5} \log{(\frac{U}ε))}$ time, where $U = \max_i f_i$. The second algorithm we provide solves a similar recurrence within a factor of $ε$ and runs in $O(n \log^2n / ε)$. The new technique introduced by our algorithm is the decomposition of the initial problem into a small (logarithmic) number of Monge optimization subproblems which we can speed up using existing techniques.

cs.DS↗